Uber customers looking to catch a ride in Pittsburgh will have the opportunity beginning this month to try the company’s new self-driving car pilot program—the first time the general public will get a taste of Uber’s autonomous automobile technology. But anyone expecting to sit back and watch Uber’s heavily modified Volvo XC90 SUVs do all of the work will probably be disappointed.
The ride-sharing company has not offered many details about what to expect during the experiment in Pittsburgh, chosen because Uber’s advanced technologies center is based there and because it gives the company a diverse set of traffic patterns, road types and weather conditions to study. Based on the current stateof self-driving technology, however, it remains to be seen whether completely autonomous maneuvering through some areas will be possible.
“Who knows where in Pittsburgh the cars will be able to be self-driving?” says Alain Kornhauser, a faculty advisor for Princeton University’s Autonomous Vehicle Engineering team and a professor of operations research and financial engineering who is not involved in the Uber test. “On the expressways they can do the intelligent cruise control, same as Mercedes and Tesla. But whether they can stop at all the red lights and stop signs by themselves? We can all wait and see.” Kornhauser cites other concerns as well. If a lane-marking line needs paint or a road sign is fading, it may be too difficult for the technology to decipher. “We know what the ends of the [autonomous technology] spectrum are,” he says. “We just don’t know how far we can push the limits of that spectrum.”
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
In Uber’s Pittsburgh experiment, two company employees (a driver and an engineer) will ride in each car along with passengers, who must first agree to take part. Unless one can afford a Tesla or is lucky enough to be invited into a Google test car—something usually reserved for dignitaries, engineers, industry analysts and the press—this will be the first time the public will have access to this type of technology. “There might be some driverless tourism as a result of Uber’s program,” says Bryant Walker Smith, an assistant professor of law and engineering at the University of South Carolina. “It’s an important first step because the public will be instrumental in shaping demand for these systems.”
Uber’s efforts to open its development of autonomous vehicles to the public do not necessarily mean the company’s technology is any closer to being self-driving than that of Tesla or Google. Uber, like its competitors, remains tight-lipped about its sensors and software, and about how it ultimately plans to eliminate the need for a safety driver and engineer in each vehicle. During the pilot project, the safety driver will monitor the system while the engineer takes notes and collects customer data. The Volvo SUVs involved in the test will have steering wheels and brake pedals so the safety driver can easily take over if needed.
No vehicles—Uber’s or others—are likely to be fully automated anytime soon, according to Smith. “The reality is these cars will be closely supervised systems because it doesn’t matter if they are 80 percent self-driving or 99 percent self-driving, you will still need a human involved for the bit that is not,” Smith says, adding that locations for use could also be limited. Companies are likely to identify places with narrow sets of conditions such as university campuses, a stretch of freeway or a downtown area as good initial targets for automated vehicles, he adds.
Uber aims to be a serious competitor in that space. This month it acquired Otto, a technology start-up co-founded by former Google employees and focused on bringing autonomous trucking to market. The acquisition gives Uber an additional 90 developers to work on its autonomous driving efforts. And it is just one of many companies, both start-up and established, testing self-driving vehicles on public roads—in every case with a safety driver at the wheel. Ford Motors, which owned Volvo from 1999 to 2010, last week announced plans to deliver a high-volume, fully autonomous vehicle for ride-sharing by 2021. The company will also supply Ford Fusion vehicles with self-driving technology to be used along with the Volvos as part of Uber’s Pittsburgh experiment. General Motors earlier this year invested $500 million in Uber competitor Lyft, which plans to begin testing a fleet of self-driving Chevrolet Bolt electric taxis within a year.
Uber’s decision to expose the public to the technology is not without inherent risks. Earlier this summer a Tesla Model S was involved in a fatal crash when its Autopilot system was activated. Tesla’s Autopilot can steer down a highway, change lanes and adjust speed in response to traffic conditions, although the company has made clear since the accident that its technology does not replace the need for a driver. The National Highway Transportation Safety Administration (NHTSA) is investigating that crash. It remains to be seen if deploying the technology was a right decision or if Tesla was too fast to market, says Marc Scribner, research fellow at the Competitive Enterprise Institute, a public policy organization. “Tesla’s rapid data collection period was able to blow other developers out of the water,” Scribner says. According to Tesla, the company has data from about 130 million miles of public road testing with Autopilot engaged, whereas Google says it has somewhere around two million road-tested miles and another three million miles per day in simulated testing.
As companies move closer to deploying vehicles for public use—without the additional company-hired safety drivers and engineers—states and developers will need to look at existing traffic and safety regulations. For example, most states actively prohibit televisions and similar devices in view of drivers, with narrow exceptions for safety and mapping screens, Scribner says. If the goal is to eventually reduce the engineer crew size of their test vehicles, legal obstacles may pose a problem. Under the current rules cars would either need to maintain a two-person crew size or maintain the driver and enable remote monitoring, Scribner says.
The NHTSA is putting the finishing touches on its new nonbinding guidance for autonomous cars, which it expects to release later this month. Based on that advice and growing data showing the real-world capabilities of these vehicles, states and developers will determine if the existing traffic and transit laws are sufficient or if changes are needed. “Over the next 10 years it will be anyone’s game,” Scribner says. “No one has commercialized the higher levels of automation, and it will be interesting to see who the first one will be.”